All posts by Nina Zumel

Working with Sessionized Data 2: Variable Selection

In our previous post in this series, we introduced sessionization, or converting log data into a form that’s suitable for analysis. We looked at basic considerations, like dealing with time, choosing an appropriate dataset for training models, and choosing appropriate (and achievable) business goals. In that previous example, we sessionized the data by considering all possible aggregations (window widths) of the data as features. Such naive sessionization can quickly lead to very wide data sets, with potentially more features than you have datums (and collinear features, as well). In this post, we will use the same example, but try to select our features more intelligently.

4203801748 f760c22c47 zIllustration: Boris Artzybasheff
photo: James Vaughan, some rights reserved

The Example Problem

Recall that you have a mobile app with both free (A) and paid (B) actions; if a customer’s tasks involve too many paid actions, they will abandon the app. Your goal is to detect when a customer is in a state when they are likely to abandon, and offer them (perhaps through an in-app ad) a more economical alternative, for example a “Pro User” subscription that allows them to do what they are currently doing at a lower rate. You don’t want to be too aggressive about showing customers this ad, because showing it to someone who doesn’t need the subscription service is likely to antagonize them (and convince them to stop using your app).

You want to build a model that predicts whether a customer will abandon the app (“exit”) within seven days. Your training set is a set of 648 customers who were present on a specific reference day (“day 0″); their activity on day 0 and the ten days previous to that (days 1 through 10), and how many days until each customer exited (Inf for customers who never exit), counting from day 0. For each day, you constructed all possible windows within those ten days, and counted the relative rates of A events and B events in each window. This gives you 132 features per row. You also have a hold-out set of 660 customers, with the same structure. You can download the wide data set used for these examples as an .rData file here. The explanation of the variable names is in the previous post in this series.

In the previous installment, we built a regularized (ridge) logistic regression model over all 132 features. This model didn’t perform too badly, but in general there is more danger of overfitting when working with very wide data sets; in addition, it is quite expensive to analyze a large number of variables with standard implementations of logistic regression. In this installment, we will look for potentially more robust and less expensive ways of analyzing this data.

Continue reading Working with Sessionized Data 2: Variable Selection

Working with Sessionized Data 1: Evaluating Hazard Models

When we teach data science we emphasize the data scientist’s responsibility to transform available data from multiple systems of record into a wide or denormalized form. In such a “ready to analyze” form each individual example gets a row of data and every fact about the example is a column. Usually transforming data into this form is a matter of performing the equivalent of a number of SQL joins (for example, Lecture 23 (“The Shape of Data”) from our paid video course Introduction to Data Science discusses this).

Doorlog

One notable exception is log data. Log data is a very thin data form where different facts about different individuals are written across many different rows. Converting log data into a ready for analysis form is called sessionizing. We are going to share a short series of articles showing important aspects of sessionizing and modeling log data. Each article will touch on one aspect of the problem in a simplified and idealized setting. In this article we will discuss the importance of dealing with time and of picking a business appropriate goal when evaluating predictive models.

For this article we are going to assume that we have sessionized our data by picking a concrete near-term goal (predicting cancellation of account or “exit” within the next 7 days) and that we have already selected variables for analysis (a number of time-lagged windows of recent log events of various types). We will use a simple model without variable selection as our first example. We will use these results to show how you examine and evaluate these types of models. In later articles we will discuss how you sessionize, how you choose examples, variable selection, and other key topics.

Continue reading Working with Sessionized Data 1: Evaluating Hazard Models

Wanted: A Perfect Scatterplot (with Marginals)

We saw this scatterplot with marginal densities the other day, in a blog post by Thomas Wiecki:

NewImage

The graph was produced in Python, using the seaborn package. Seaborn calls it a “jointplot;” it’s called a “scatterhist” in Matlab, apparently. The seaborn version also shows the strength of the linear relationship between the x and y variables. Nice.

I like this plot a lot, but we’re mostly an R shop here at Win-Vector. So we asked: can we make this plot in ggplot2? Natively, ggplot2 can add rugs to a scatterplot, but doesn’t immediately offer marginals, as above.

However, you can use Dean Attali’s ggExtra package. Here’s an example using the same data as the seaborn jointplot above; you can download the dataset here.

library(ggplot2)
library(ggExtra)
frm = read.csv("tips.csv")

plot_center = ggplot(frm, aes(x=total_bill,y=tip)) + 
  geom_point() +
  geom_smooth(method="lm")

# default: type="density"
ggMarginal(plot_center, type="histogram")

I didn’t bother to add the internal annotation for the goodness of the linear fit, though I could.

PltggMarginal

The ggMarginal() function goes to heroic effort to line up the coordinate axes of all the graphs, and is probably the best way to do a scatterplot-plus-marginals in ggplot (you can also do it in base graphics, of course). Still, we were curious how close we could get to the seaborn version: marginal density and histograms together, along with annotations. Below is our version of the graph; we report the linear fit’s R-squared, rather than the Pearson correlation.

# our own (very beta) plot package: details later
library(WVPlots)
frm = read.csv("tips.csv")

ScatterHist(frm, "total_bill", "tip",
            smoothmethod="lm",
            annot_size=3,
            title="Tips vs. Total Bill")

PlotPkg

You can see that (at the moment) we’ve resorted to padding the axis labels with underbars to force the x-coordinates of the top marginal plot and the scatterplot to align; white space gets trimmed. This is profoundly unsatisfying, and less robust than the ggMarginal version. If you’re curious, the code is here. It relies on some functions in the file sharedFunctions.R in the same repository. Our more general version will do either a linear or lowess/spline smooth, and you can also adjust the histogram and density plot parameters.

Thanks to Slawa Rokicki’s excellent ggplot2: Cheatsheet for Visualizing Distributions for our basic approach. Check out the graph at the bottom of her post — and while you’re at it, check out the rest of her blog too.

Does Balancing Classes Improve Classifier Performance?

It’s a folk theorem I sometimes hear from colleagues and clients: that you must balance the class prevalence before training a classifier. Certainly, I believe that classification tends to be easier when the classes are nearly balanced, especially when the class you are actually interested in is the rarer one. But I have always been skeptical of the claim that artificially balancing the classes (through resampling, for instance) always helps, when the model is to be run on a population with the native class prevalences.

On the other hand, there are situations where balancing the classes, or at least enriching the prevalence of the rarer class, might be necessary, if not desirable. Fraud detection, anomaly detection, or other situations where positive examples are hard to get, can fall into this case. In this situation, I’ve suspected (without proof) that SVM would perform well, since the formulation of hard-margin SVM is pretty much distribution-free. Intuitively speaking, if both classes are far away from the margin, then it shouldn’t matter whether the rare class is 10% or 49% of the population. In the soft-margin case, of course, distribution starts to matter again, but perhaps not as strongly as with other classifiers like logistic regression, which explicitly encodes the distribution of the training data.

So let’s run a small experiment to investigate this question.

Continue reading Does Balancing Classes Improve Classifier Performance?

Random Test/Train Split is not Always Enough

Most data science projects are well served by a random test/train split. In our book Practical Data Science with R we strongly advise preparing data and including enough variables so that data is exchangeable, and scoring classifiers using a random test/train split.

With enough data and a big enough arsenal of methods, it’s relatively easy to find a classifier that looks good; the trick is finding one that is good. What many data science practitioners (and consumers) don’t seem to remember is that when evaluating a model, a random test/train split may not always be enough.

Continue reading Random Test/Train Split is not Always Enough

The Geometry of Classifiers

As John mentioned in his last post, we have been quite interested in the recent study by Fernandez-Delgado, et.al., “Do we Need Hundreds of Classifiers to Solve Real World Classification Problems?” (the “DWN study” for short), which evaluated 179 popular implementations of common classification algorithms over 120 or so data sets, mostly from the UCI Machine Learning Repository. For fun, we decided to do a follow-up study, using their data and several classifier implementations from scikit-learn, the Python machine learning library. We were interested not just in classifier accuracy, but also in seeing if there is a “geometry” of classifiers: which classifiers produce predictions patterns that look similar to each other, and which classifiers produce predictions that are quite different? To examine these questions, we put together a Shiny app to interactively explore how the relative behavior of classifiers changes for different types of data sets.

Continue reading The Geometry of Classifiers

Estimating Generalization Error with the PRESS statistic

As we’ve mentioned on previous occasions, one of the defining characteristics of data science is the emphasis on the availability of “large” data sets, which we define as “enough data that statistical efficiency is not a concern” (note that a “large” data set need not be “big data,” however you choose to define it). In particular, we advocate the use of hold-out data to evaluate the performance of models.

There is one caveat: if you are evaluating a series of models to pick the best (and you usually are), then a single hold-out set is strictly speaking not enough. Hastie, et.al, say it best:

Ideally, the test set should be kept in a “vault,” and be brought out only at the end of the data analysis. Suppose instead that we use the test-set repeatedly, choosing the model with smallest test-set error. Then the test set error of the final chosen model will underestimate the true test error, sometimes substantially.

— Hastie, Tibshirani and Friedman, The Elements of Statistical Learning, 2nd edition.

The ideal way to select a model from a set of candidates (or set parameters for a model, for example the regularization constant) is to use a training set to train the model(s), a calibration set to select the model or choose parameters, and a test set to estimate the generalization error of the final model.

In many situations, breaking your data into three sets may not be practical: you may not have very much data, or the the phenomena you’re interested in are rare enough that you need a lot of data to detect them. In those cases, you will need more statistically efficient estimates for generalization error or goodness-of-fit. In this article, we look at the PRESS statistic, and how to use it to estimate generalization error and choose between models.

Continue reading Estimating Generalization Error with the PRESS statistic

Vtreat: designing a package for variable treatment

When you apply machine learning algorithms on a regular basis, on a wide variety of data sets, you find that certain data issues come up again and again:

  • Missing values (NA or blanks)
  • Problematic numerical values (Inf, NaN, sentinel values like 999999999 or -1)
  • Valid categorical levels that don’t appear in the training data (especially when there are rare levels, or a large number of levels)
  • Invalid values

Of course, you should examine the data to understand the nature of the data issues: are the missing values missing at random, or are they systematic? What are the valid ranges for the numerical data? Are there sentinel values, what are they, and what do they mean? What are the valid values for text fields? Do we know all the valid values for a categorical variable, and are there any missing? Is there any principled way to roll up category levels? In the end though, the steps you take to deal with these issues will often be the same from data set to data set, so having a package of ready-to-go functions for data treatment is useful. In this article, we will discuss some of our usual data treatment procedures, and describe a prototype R package that implements them.

Continue reading Vtreat: designing a package for variable treatment

Trimming the Fat from glm() Models in R

One of the attractive aspects of logistic regression models (and linear models in general) is their compactness: the size of the model grows in the number of coefficients, not in the size of the training data. With R, though, glm models are not so concise; we noticed this to our dismay when we tried to automate fitting a moderate number of models (about 500 models, with on the order of 50 coefficients) to data sets of moderate size (several tens of thousands of rows). A workspace save of the models alone was in the tens of gigabytes! How is this possible? We decided to find out.

As many R users know (but often forget), a glm model object carries a copy of its training data by default. You can use the settings y=FALSE and model=FALSE to turn this off.

set.seed(2325235)


# Set up a synthetic classification problem of a given size
# and two variables: one numeric, one categorical
# (two levels).
synthFrame = function(nrows) {
   d = data.frame(xN=rnorm(nrows),
      xC=sample(c('a','b'),size=nrows,replace=TRUE))
   d$y = (d$xN + ifelse(d$xC=='a',0.2,-0.2) + rnorm(nrows))>0.5
   d
}


# first show that model=F and y=F help reduce model size

dTrain = synthFrame(1000)
model1 = glm(y~xN+xC,data=dTrain,family=binomial(link='logit'))
model2 = glm(y~xN+xC,data=dTrain,family=binomial(link='logit'),
             y=FALSE)
model3 = glm(y~xN+xC,data=dTrain,family=binomial(link='logit'),
              y=FALSE, model=FALSE)

#
# Estimate the object's size as the size of its serialization
#
length(serialize(model1, NULL))
# [1] 225251
length(serialize(model2, NULL))
# [1] 206341
length(serialize(model3, NULL))
# [1] 189562

dTest = synthFrame(100)
p1 = predict(model1, newdata=dTest, type='response')
p2 = predict(model2, newdata=dTest, type='response')
p3 = predict(model3, newdata=dTest, type='response')
sum(abs(p1-p2))
# [1] 0
sum(abs(p1-p3))
# [1] 0

Continue reading Trimming the Fat from glm() Models in R

Bandit Formulations for A/B Tests: Some Intuition

Controlled experiments embody the best scientific design for establishing a causal relationship between changes and their influence on user-observable behavior.

— Kohavi, Henne, Sommerfeld, “Practical Guide to Controlled Experiments on the Web” (2007)

A/B tests are one of the simplest ways of running controlled experiments to evaluate the efficacy of a proposed improvement (a new medicine, compared to an old one; a promotional campaign; a change to a website). To run an A/B test, you split your population into a control group (let’s call them “A”) and a treatment group (“B”). The A group gets the “old” protocol, the B group gets the proposed improvement, and you collect data on the outcome that you are trying to achieve: the rate that patients are cured; the amount of money customers spend; the rate at which people who come to your website actually complete a transaction. In the traditional formulation of A/B tests, you measure the outcomes for the A and B groups, determine which is better (if either), and whether or not the difference observed is statistically significant. This leads to questions of test size: how big a population do you need to get reliably detect a difference to the desired statistical significance? And to answer that question, you need to know how big a difference (effect size) matters to you.

The irony is that to detect small differences accurately you need a larger population size, even though in many cases, if the difference is small, picking the wrong answer matters less. It can be easy to lose sight of that observation in the struggle to determine correct experiment sizes.

There is an alternative formulation for A/B tests that is especially suitable for online situations, and that explicitly takes the above observation into account: the so-called multi-armed bandit problem. Imagine that you are in a casino, faced with K slot machines (which used to be called “one-armed bandits” because they had a lever that you pulled to play (the “arm”) — and they pretty much rob you of all your money). Each of the slot machines pays off at a different (unknown) rate. You want to figure out which of the machines pays off at the highest rate, then switch to that one — but you don’t want to lose too much money to the suboptimal slot machines while doing so. What’s the best strategy?

NewImage

The “pulling one lever at a time” formulation isn’t a bad way of thinking about online transactions (as opposed to drug trials); you can imagine all your customers arriving at your site sequentially, and being sent to bandit A or bandit B according to some strategy. Note also, that if the best bandit and the second-best bandit have very similar payoff rates, then settling on the second best bandit, while not optimal, isn’t necessarily that bad a strategy. You lose winnings — but not much.

Traditionally, bandit games are infinitely long, so analysis of bandit strategies is asymptotic. The idea is that you test less as the game continues — but the testing stage can go on for a very long time (often interleaved with periods of pure exploitation, or playing the best bandit). This infinite-game assumption isn’t always tenable for A/B tests — for one thing, the world changes; for another, testing is not necessarily without cost. We’ll look at finite games below.

Continue reading Bandit Formulations for A/B Tests: Some Intuition